Meta launches Muse Code, a terminal AI coding agent

Meta Superintelligence Labs released Muse Code in beta, a terminal-based coding agent powered by a new model called Muse Spark 1.2. The agent plans changes, writes code, and validates results across large repositories, and its async background agents stay active for a whole session instead of spawning per task. A local append-only event log makes the runtime replay-exact and restart-safe after a crash.

Why it matters: This puts Meta squarely into the crowded agentic coding tool market alongside Claude Code, OpenAI Codex, and Cursor, signaling every major lab now treats terminal coding agents as core infrastructure rather than a niche product. Co-training Muse Spark 1.2 with the harness mirrors the approach Anthropic and OpenAI have taken, suggesting model-harness co-design is becoming the standard playbook for long-horizon coding work.